Computational Architecture and Method for a Time-Varying Control System

1. A system for controlling an enterprise that produces a plurality of products by at least one production process, the system comprising: computing hardware, including at least one processor, data storage, and input/output facilities operatively coupled to the at least one production process of the enterprise, the data storage containing instructions that, when executed by the at least one processor, cause the computing hardware to implement: a main processing engine configured to generate a control vector corresponding to a target allocation of resources between the plurality of products to be produced by the enterprise, the control vector being applied to adjust operation of the at least one production process for each of the products; an actual revenue measurement engine configured to obtain data representing actual revenue generated by sales of each of the plurality of products in a market during a first time period; an actual cost measurement engine configured to obtain data representing actual cost attributed to each of the plurality of products during the first time period; a retained earnings computation engine configured to produce an observation vector containing elements in a data structure representing actual retained earnings attributed to each of the plurality of products during the first time period, the actual retained earnings being based on the data representing the actual revenue and on the data representing the actual cost; a matrix generator engine configured to generate a dynamic matrix, a control matrix, a cost matrix, and an observation matrix, wherein: the dynamic matrix contains elements arranged in a data structure that represents a dynamic relationship between a current set of revenue values and a predicted set of revenue values at a second time period subsequent to the first time period; the control matrix contains elements arranged in a data structure that represents a dynamic relationship between the control vector and the actual revenue corresponding to each product depending on the control vector; the cost matrix contains elements arranged in a data structure that represents a dynamic relationship between variations in the market and actual revenue corresponding to each product depending on the variations in the market; the observation matrix contains elements arranged in a data structure that represents a dynamic relationship between the actual revenue corresponding to each product and the actual retained earnings; the system further comprising an estimator engine configured to revise the dynamic matrix, the control matrix, the cost matrix, and the observation matrix for the second time period to produce predicted values for each corresponding matrix, the predicted values being based on corresponding historical values from at least one prior time period, wherein a degree of accuracy of the predicted values and computational loading of the computing hardware in implementation of the estimator engine increase with an increase in quantity of the at least one prior time period, and wherein the quantity of the at least one prior time period is limited in accordance with a desired level of accuracy of the predicted values.

2. The system of claim 1 , wherein the control vector comprises a vector having a number of elements equal to a quantity of the plurality of products.

3. The system of claim 2 , wherein a sum of elements that make up the control vector is equal to one.

4. The system of claim 1 , wherein the estimator engine is configured to apply a set of forecasting algorithms to each of the vectorized dynamic, control, cost, and observation matrices, each forecasting algorithm being associated with a set of prediction errors, and wherein the estimator engine further selects one of the set of forecasting algorithms that is associated with a first prediction error that is smaller than the others of the set of prediction errors.

5. The system of claim 4 , wherein predictions of each of the vectorized dynamic, control, cost, and observation matrices are generated using exponential smoothing.

6. The system of claim 4 , wherein predictions of each of the vectorized dynamic, control, cost, and observation matrices are generated using a Holt-Winters method.

7. The system of claim 1 , wherein the control vector is generated such that an objective function is minimized.

8. The system of claim 1 , wherein the control vector is generated by applying stabilization criteria to stabilize revenues for each of the products over a series of time periods.

9. The system of claim 1 , wherein the dynamic matrix is estimated based on an application of an analytical relationship between projected dynamic matrix values and measured historical values following from a dynamic equation with linear control and cost laws.

10. A method for controlling an enterprise that produces a plurality of products by at least one production process, the method comprising: generating, by a computing system, a control vector corresponding to a target allocation of resources between the plurality of products to be produced by the enterprise, the control vector being applied to adjust operation of the at least one production process for each of the products; obtaining, by the computing system, data representing actual revenue generated by sales of each of the plurality of products in a market during a first time period; obtaining, by the computing system, data representing actual cost attributed to each of the plurality of products during the first time period; producing, by the computing system, an observation vector containing elements in a data structure representing actual retained earnings attributed to each of the plurality of products during the first time period, the actual retained earnings being based on the data representing the actual revenue and on the data representing the actual cost; generating, by the computing system, a dynamic matrix, a control matrix, a cost matrix, and an observation matrix, wherein: the dynamic matrix contains elements arranged in a data structure that represents a dynamic relationship between a current set of revenue values and a predicted set of revenue values at a second time period subsequent to the first time period; the control matrix contains elements arranged in a data structure that represents a dynamic relationship between the control vector and the actual revenue corresponding to each product depending on the control vector; the cost matrix contains elements arranged in a data structure that represents a dynamic relationship between variations in the market and actual revenue corresponding to each product depending on the variations in the market; the observation matrix contains elements arranged in a data structure that represents a dynamic relationship between the actual revenue corresponding to each product and the actual retained earnings; the method further comprising revising, by the computing system, the dynamic matrix, the control matrix, the cost matrix, and the observation matrix for the second time period to produce predicted values for each corresponding matrix, the predicted values being based on corresponding historical values from at least one prior time period, wherein a degree of accuracy of the predicted values and computational loading of the computing system in revising the dynamic matrix, the control matrix, the cost matrix, and the observation matrix for the second time period increase with an increase in quantity of the at least one prior time period, and wherein the quantity of the at least one prior time period is limited in accordance with a desired level of accuracy of the predicted values.

11. The method of claim 10 , wherein the control vector comprises a vector having a number of elements equal to a quantity of the plurality of products.

12. The method of claim 11 , wherein a sum of elements that make up the control vector is equal to one.

13. The method of claim 10 , in the revising, the computer system applies a plurality of forecasting algorithms to each of the vectorized dynamic, control, cost, and observation matrices, each forecasting algorithm being associated with a set of prediction errors, and wherein the computer system further selects one of the set of forecasting algorithms that is associated with a first prediction error that is smaller than the others of the set of prediction errors.

14. The method of claim 10 , wherein the control vector is generated such that an objective function is minimized.

15. The method of claim 10 , wherein the control vector is generated by applying stabilization criteria to reduce fluctuations in revenue.

16. The method of claim 10 , wherein the dynamic matrix is estimated based on an application of an analytical relationship between projected dynamic matrix values and measured historical values following from a dynamic equation with linear control and cost laws.

17. A logistics system for an enterprise having a resource allocation control system, the logistics system comprising: a data interface with a business system associated with one or more product lines; means for generating a control vector corresponding to a target allocation of resources between the plurality of products to be produced by the enterprise, the control vector being applied to adjust operation of the at least one production process for each of the products; means for obtaining data representing actual revenue generated by sales of each of the plurality of products in a market during a first time period; means for obtaining data representing actual cost attributed to each of the plurality of products during the first time period; means for producing an observation vector containing elements in a data structure representing actual retained earnings attributed to each of the plurality of products during the first time period, the actual retained earnings being based on the data representing the actual revenue and on the data representing the actual cost; means for generating, by the computing system, a dynamic matrix, a control matrix, a cost matrix, and an observation matrix, wherein: the dynamic matrix contains elements arranged in a data structure that represents a dynamic relationship between a current set of revenue values and a predicted set of revenue values at a second time period subsequent to the first time period; the control matrix contains elements arranged in a data structure that represents a dynamic relationship between the control vector and the actual revenue corresponding to each product depending on the control vector; the cost matrix contains elements arranged in a data structure that represents a dynamic relationship between variations in the market and actual revenue corresponding to each product depending on the variations in the market; the observation matrix contains elements arranged in a data structure that represents a dynamic relationship between the actual revenue corresponding to each product and the actual retained earnings; means for revising the dynamic matrix, the control matrix, the cost matrix, and the observation matrix for the second time period to produce predicted values for each corresponding matrix, the predicted values being based on corresponding historical values from at least one prior time period, wherein a degree of accuracy of the predicted values and computational loading of the logistics system in implementing the means for revising the dynamic matrix, the control matrix, the cost matrix, and the observation matrix for the second time period increase with an increase in quantity of the at least one prior time period, and wherein the quantity of the at least one prior time period is limited in accordance with a desired level of accuracy of the predicted values.

18. The logistics system of claim 17 , wherein the control vector is generated such that an objective function is minimized.

19. The logistics system of claim 17 , wherein the control vector is generated by applying stabilization criteria to stabilize revenues for each of the products over a series of time periods.

20. The logistics system of claim 17 , wherein the dynamic matrix is estimated based on an application of an analytical relationship between projected dynamic matrix values and measured historical values following from a dynamic equation with linear control and cost laws.